Insurance robo-advisors automate pricing, underwriting, and policy servicing through machine-learning models. While they promise efficiency and objectivity, their opacity amplifies systemic risks: data mapping errors inflate premiums, model drift voids coverage, and biased proxies trigger unlawful discrimination. Private law remedies may not always be sufficient. Privity blocks claims against upstream software vendors; exclusion clauses cap recovery; proving fault inside proprietary code is practically impossible. The revised Product Liability Directive (PLD) seeks to plug this gap by classifying “software, digital manufacturing files and AI systems” as products and imposing strict liability on manufacturers, importers and white-labelling insurers. However, essential uncertainties remain: the boundary between an adaptive algorithm and a ‘defect’ is unclear; pure economic loss is still only indirectly covered; and multi-tier vendor chains complicate the allocation of liability. By contrasting contracting-based and product-based enforcement, this chapter argues that PLD closes much, but not all, of the robo-advisor liability gap, effectively shifting rather than eliminating residual legal uncertainty.